RNP2 of RNA Recognition Motif 1 Plays a Central Role in the Aberrant Modification of TDP-43

Phosphorylated and truncated TAR DNA-binding protein-43 (TDP-43) is a major component of ubiquitinated cytoplasmic inclusions in neuronal and glial cells of two TDP-43 proteinopathies, amyotrophic lateral sclerosis and frontotemporal lobar degeneration. Modifications of TDP-43 are thus considered to play an important role in the pathogenesis of TDP-43 proteinopathies. However, both the initial cause of these abnormal modifications and the TDP-43 region responsible for its aggregation remain uncertain. Here we report that the 32 kDa C-terminal fragment of TDP-43, which lacks the RNP2 motif of RNA binding motif 1 (RRM1), formed aggregates in cultured cells, and that similar phenotypes were obtained when the RNP2 motif was either deleted from or mutated in full-length TDP-43. These aggregations were ubiquitinated, phosphorylated and truncated, and sequestered the 25 kDa C-terminal TDP-43 fragment seen in the neurons of TDP-43 proteinopathy patients. In addition, incubation with RNase decreased the solubility of TDP-43 in cell lysates. These findings suggest that the RNP2 motif of RRM1 plays a substantial role in pathological TDP-43 modifications and that it is possible that disruption of RNA binding may underlie the process of TDP-43 aggregation.


Introduction
Amyotrophic lateral sclerosis (ALS) and certain forms of frontotemporal lobar degeneration (FTLD) with ubiquitin-positive but tau-negative inclusions have been considered to be a single disease spectrum of TAR DNA-binding protein 43 (TDP-43) proteinopathy since 2006, when this protein was reported to be a major component of ubiquitin-positive inclusions in the affected neuronal and glial cells of both ALS and FTLD [1][2][3]. The identification of missense mutations of TARDBP, the gene encoding TDP-43, in familial and sporadic ALS and/or FTLD patients further confirmed the importance of this molecule in the pathogenesis of TDP-43 proteinopathies [4][5][6][7].
Although TDP-43 normally localizes to the nucleus, it is distributed from nucleus to cytoplasm or neurite and forms aggregates that mainly consist of C-terminal fragments (CTFs) in the affected neurons of TDP-43 proteinopathy patients. In addition, aberrantly aggregated TDP-43 is hyperphosphorylated at multiple C-terminal sites [8]. The fact that most TDP-43 proteinopathy cases are sporadic suggests that exogenous factors induce the post-translational modifications of TDP-43 that are seen in the disease.
These findings provide us with a clue for elucidating the mechanism of these modifications. Nevertheless, both the initial cause of these abnormal modifications and the region of TDP-43 responsible for its aggregation remain unknown. Here we report that RNP2 in RNA binding motif 1 (RRM1) plays a substantial role in the pathological TDP-43 modifications that are seen in TDP-43 proteinopathies.

Cell Culture and Treatment
Mouse NSC34 motor neuron-like cells (a kind gift of N.R. Cashman, University of British Columbia, Vancouver, Canada) [23] and human embryonic kidney 293 (HEK293) cells were cultured in a humidified atmosphere of 95% air/5% CO2 in a 37uC incubator in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS). In both NSC34 and HEK293 cells, the transfections of the intended plasmids were performed using Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. Before performing subsequent experiments, the cells were incubated for 48 h after transfection.

DNA Constructs
Human wild-type TDP-43 (WT-TDP-43) (accession number NM 007375) cDNA was amplified by PCR from human spinal cord cDNA as previously described [11]. The PCR product was cloned into the pENTR/D-TOPO vector (Invitrogen). For the TDP-43 truncated fragments, amplified PCR products (for primers see Table S1) from the WT-TDP-43 vector were cloned into the pENTR/D-TOPO vector. For delta RNP2 TDP-43 (DRNP2), mutated RNP2 TDP-43 (mtRNP2), mutated RNP1 TDP-43 (mRNP1), delta RRM1 TDP-43 (DRRM1), and delta nuclear localization signal (NLS) TDP-43 (dNLS) vectors, primers containing mutant substitutions (Table S2) were used to mutagenize WT-TDP-43 (KOD-Plus-Mutagenesis kit; Toyobo). The entry vectors of each mutant TDP-43 vector were transferred into either a pcDNA6.2/N-EmGFP-DEST or pcDNA3.1/nV5-DEST vector using the Gateway LR Clonase II enzyme mix (Invitrogen). The sequences of all constructs were verified using the CEQ 8000 genetic analysis system (Beckman Coulter). The collection of autopsied human tissue and its use for this study were approved by the Ethics Committee of Nagoya University Graduate School of Medicine, and written informed consent was obtained from the patients' next-of-kin. The experimental procedure involving the human subject was conducted in conformance with the principles expressed in the Declaration of Helsinki.

Immunoblot Analysis
We used NE-PER Nuclear Cytoplasmic Reagents (Thermo Fisher Scientific) for the analysis of the cytoplasmic/nuclear ratio. For analysis of protein solubility, cells cultured in 10-cm plates were lysed in 1 ml RIPA buffer (Thermo Fisher Scientific). Lysates were sonicated and centrifuged at 100,000 g for 15 min. To prevent carryover, the pellets were washed with RIPA buffer, followed by sonication and centrifugation. RIPA-insoluble pellets were lysed in 100 ml urea buffer (7 M urea, 2 M thiourea, 4% CHAPS, 30 mM Tris, pH 8.5), sonicated, and centrifuged at 100,000 g for 15 min.
For the counting of inclusion-bearing cells, we randomly selected 100 transfected cells from three separate experiments. The colocalization coefficient, which reflects the fraction of green pixels that are also positive for red pixels, was calculated using the Zeiss LSM software. We calculated the colocalization coefficient by randomly selected 10 fields from three separate experiments. To obtain images for calculating the colocalization coefficient, the settings of the confocal microscopy and the threshold of positive/ negative fluorescence was fixed within each experiment.

Immunoprecipitation
Transfected HEK293 cells were washed with PBS and lysed in immunoprecipitation buffer (Thermo Fisher Scientific). After sonication on ice, the samples were agitated for 30 min at 4uC. The samples were centrifuged and supernatants were incubated with magnetic beads: anti-V5 magnetic beads (MBL), anti-GFP magnetic beads (MBL), and anti-ubiquitin magnetic beads (MBL). Samples were rotated overnight at 4uC. Immunoprecipitates were separated by SDS-PAGE (5%-20% gradient gel). Western blotting was performed using anti-V5-HRP antibody (MBL) and anti-GFP-HRP antibody (MBL).

Ribonucleoprotein Immunoprecipitation
Ribonucleoprotein (RNP) immunoprecipitation was performed using a RIP assay kit (MBL), according to the manufacturer's instructions. RNA concentrations were measured with a Nanodrop (Thermo Fisher Scientific). Electrophoresis of precipitated RNA was performed with a Bioanalyzer (Agilent Technologies) according to the manufacturer's instructions. For analysis of neurofilament light chain (hNFL) mRNA 39UTR content, RNA obtained from immunoprecipitates was reverse transcribed into first-strand cDNA using SuperScript II reverse transcriptase (Invitrogen) and a PCR was performed with the following primers: ACCAACCAGTTGAGTTCCAGAT (forward) and GAAT-GATTCACATTGCCGTAGA (reverse).

Effect of RNase on TDP-43 Solubility
For analysis of protein solubility with or without RNase, HEK293 cells cultured in 10-cm plates were lysed in 1 ml of Trissaline (TS) buffer (50 mM Tris-HCl buffer, pH 7.5, 0.15 M NaCl, 5 mM EDTA, protein phosphatase inhibitors, and a protease inhibitor cocktail). Lysates were sonicated and then divided into two samples. RNase A (10 mg/ml) was added to one of the samples. Samples incubated for 0 and 15 h at 4uC were centrifuged at 100,000 g for 15 min. To prevent carryover, the pellets were washed with TS buffer, followed by sonication and centrifugation. TS-insoluble pellets were lysed in 1 ml of Triton X-100 (TX) buffer (TS buffer containing 1% Triton X-100), sonicated, and centrifuged at 100,000 g for 15 min. The pellets were washed with TX buffer, followed by sonication and centrifugation. TX-insoluble pellets were lysed in 500 ml of Sarkosyl (Sar) buffer (TS buffer containing 1% Sarkosyl), sonicated and centrifuged at 100,000 g for 15 min. Sar-insoluble pellets were lysed in 100 ml of urea buffer.

Statistical Analyses
Statistical analyses were performed using GraphPad Prism software (GraphPad software inc.). Biochemical data were statistically analyzed using a Student's t-test or one-factor factorial ANOVA followed by Tukey post hoc tests. A p value of 0.05 or less was considered to be statistically significant.

Intracellular Localization of CTFs of TDP-43 in NSC34 Cells
To identify the region of TDP-43 that is responsible for the pathological modification of this protein, we created various TDP-43 mutants and investigated their intracellular localizations. In particular, given that the CTFs of TDP-43 form aggregates in the cytoplasm of affected neurons, we focused on the mutant TDP-43 in which the NLS is disrupted (dNLS) and on TDP-43 CTFs: 35 kDa (TDP35), 32 kDa (TDP32), and 25 kDa (TDP25) fragments (Fig. 1A). We examined TDP32, which does not contain RNP2 motif (aa 106-111), to access the effect of RNAbinding on the pathological modification, since RNP2 motif is indispensable for RNA-binding of RRM1 [24]. Each form of TDP-43 was transfected into NSC34 cells for immunocytochemical analysis and into HEK293 cells for the analysis of their localization using fractionated immunoblots.
Although wild-type TDP-43 showed a nuclear-dominant distribution, the dNLS mutant localized to the cytosol more preferentially than wild-type TDP-43 (Fig. 1B). The CTFs of TDP-43, all of which lack the NLS, also showed a cytosolic localization. In particular, the short CTFs, TDP32 and TDP25, had a strong propensity to distribute in the cytosol. We also found that the total soluble protein levels of TDP32 and TDP25 were less than that of TDP35, suggesting that these two CTFs may be insoluble or rapidly degraded.

Ubiquitination of TDP-43 CTFs
The relatively low steady-state levels of TDP32 and TDP25 prompted us to explore the possibility that these CTFs form aggregates. Given that the expression levels of TDP25 are substantially lower than other forms of TDP-43, we first focused on dNLS, TDP35, and TDP32. In immunocytochemical analysis using anti-ubiquitin and -V5 antibodies, wild-type TDP-43 chiefly localized to the nucleus, but a substantial amount of dNLS and TDP35 distributed to the cytosol ( Fig. 2A). TDP32 also showed a cytosolic localization but formed aggregates that were stained with anti-ubiquitin antibody ( Fig. 2A, B).
In contrast, ubiquitin-positive aggregates were virtually undetectable in the cells expressing TDP35 (Fig. 2B). Since TDP-43 is reported to form RNA-containing structures like stress granules, we examined the relationship between TDP-43 CTFs and T-cellrestricted intracellular antigen 1-related (TIAR) protein, a marker of stress granules. Although TDP35 occasionally colocalized with TIAR, the inclusions of TDP32 were distinct from anti-TIARstained RNA granules (Fig. S1).
To quantitatively analyze the relationship between ubiquitin and TDP-43 CTFs, we calculated the colocalization coefficient of ubiquitin and V5 immunofluorescence. We found that the colocalization coefficient of ubiquitin and V5 was significantly higher in TDP32-expressing cells than in those expressing TDP35 (p,0.001; Fig. S2). We also confirmed using immunoprecipitation the differential ubiquitination of TDP35 and TDP32. Anti- ubiquitin immunoprecipitates were only detected in the cells expressing TDP32, although the steady-state levels of TDP32 were far lower than those of TDP35 (Fig. 2C). Immunoblotting of anti-V5 immunoprecipitates also showed that the ubiquitin-positive smear was denser in the cells bearing TDP32 than in those expressing TDP35 (Fig. 2D). Together, these results suggest that TDP32, but not TDP35, forms ubiquitin-positive aggregates.

Insolubilization and Phosphorylation of the 32 kDa CTF of TDP-43
Like ubiquitination, insolubilization and phosphorylation are characteristics of TDP-43 proteinopathies. Therefore, we investigated TDP-43 CTF solubility and phosphorylation, finding that although the cytosolic aggregates of TDP32 were well stained with anti-phosphorylated TDP-43 (pTDP-43) antibody, the other forms showed no detectable pTDP-43-containing aggregates (Fig. 3A). The quantitative analysis also confirmed that the number of pTDP-43-positive aggregates was significantly higher in the cells expressing TDP32 than in those transfected with the TDP35 vector (p,0.01; Fig. 3B). The colocalization coefficient of TDP-43 CTFs and phosphorylated TDP-43 was also significantly higher in the cells expressing TDP32 than in those with TDP35 (p,0.001; Fig. S3). In the immunoblots performed using RIPA lysis buffer, the amount of insoluble TDP32 was higher than that of TDP35, whereas most of the TDP35 was solubilized by this buffer (Fig. 3C). TDP32 in the RIPA-insoluble fraction was also detected with an anti-pTDP-43 antibody (Fig. 3D).

Disruption of the RNP2 Motif in TDP-43 Leads to Ubiquitin-positive Aggregate Formation
The phenotype of TDP32 was distinctly different from that of TDP35, although both of which lack the NLS, span amino acids 85-414 and 112-414, respectively (Fig. 1A). Although these two CTFs, and share a common structure with regard to the RRM2 and glycine-rich domains. The critical difference between these CTFs is the RNP2 (aa 106-111), the RNA binding motif at its N-terminal portion, which was included in TDP35 but not in TDP32. The contrast between TDP35 and TDP32 regarding modifications suggests that RNP2 motif is responsible for the induction of critical changes such as ubiquitination, phosphorylation, and insolubilization.
To investigate the role of the RNP2 motif in the modification of TDP-43, we created two defective mutants, DRNP2 and mtRNP2 (Fig. 4A). While DRNP2 has no RNP2 motif, mtRNP2 contains the mutated RNP2 in which leucine residues were changed to aspartic acid, as previously reported [24]. The NSC34 cells overexpressing these mutants bear ubiquitin-positive inclusions in both the nucleus and cytosol (Fig. 4B, C). In immunocytochemical analyses, the inclusions of DRNP2 and mtRNP2 were distinct from RNA granules labeled with the anti-TIAR antibody (Fig. S4). The colocalization coefficient of mtRNP2 and ubiquitin was significantly higher than that of wild-type TDP-43 and ubiquitin (p,0.001; Fig. S5).
Immunoprecipitation analyses showed that mtRNP2 inclusions were strongly ubiquitinated in comparison with those of wild-type TDP-43 (Fig. 4D, E). Inclusions of DRNP2 and mtRNP2 were also immunoreactive to the anti-pTDP-43 antibody (Fig. 5A, B). The colocalization coefficient of mtRNP2 and pTDP-43 was significantly higher than that of wild-type TDP-43 and pTDP-43 (p,0.05; Fig. S6). The amount of mtRNP2 in the RIPA-insoluble fraction was higher than that of wild-type TDP-43 (Fig. 5C). Additionally, mtRNP2 was more phosphorylated than wild-type TDP-43 (Fig. 5D). These results further demonstrate the features of TDP32, supporting the view that RNP2 has a protective role for the pathological modification of TDP-43. Since V5 tag possibly has a certain effect on these modifications, we also assessed the biological properties of non-tagged mtRNP2 TDP-43. The results showed that mtRNP2 without a tag became also insoluble and hyperphosphorylated (Fig. S7A). In addition, we investigated TDP-43 with disrupted RNP1 motif (mtRNP1) in RRM1, which is also responsible for RNA binding of TDP-43. The mtRNP1 was insoluble and phosphorylated compared with wild-type TDP-43 (Fig. S7B).

Biological Features of the 25 kDa CTF of TDP-43
Small CTFs of 18-26 kDa accumulate in the cytosol of affected neurons of TDP-43 proteinopathies [8,25]. However, in the present cellular study, the 25 kDa CTF of TDP-43 (TDP25) was soluble (Fig. 6A) and scarcely formed ubiquitin-positive aggregates (Fig. 6B). These findings indicate that TDP25 is not insoluble when simply overexpressed in cultured cells. Given that TDP25 contains the complete RRM2 and C-terminal domains of TDP-43, our results suggest that these domains of TDP-43 do not by themselves play an essential role in aggregation. To confirm this hypothesis, we assessed the phenotype of C-terminal domainlacking mtRNP2 mutants: mtRNP2 (1-273), which has RRM1 and RRM2; mtRNP2 (1-185), which contains RRM1 but not RRM2; and TDP-43 (1-105), which lacks both RRM1 and RRM2. The results showed that both mtRNP2 (1-273) and mtRNP2 (1-185) formed aggregates, whereas TDP-43 (1-105) lacking both RRMs diffusely located in the nucleus without aggregation (Fig. S8A, B). In addition, DRRM1 of TDP-43 did not apparently form aggregates (Fig. S8A, B). Although DRRM1 showed a punctate nuclear localization, it was not insoluble or phosphorylated in the immunoblots (Fig. S8C). Taken together, the disrupted RNP2 in conjunction with the remaining RRM1 is likely to be necessary for the aggregation of TDP-43.
Next we investigated whether the 18-26 kDa TDP-43 fragments are included in the aggregates of the CTFs we created. Following prolonged exposure of the immunoblots of TDP-43 mutants, we found that 25 kDa and shorter fragments were detected in the insoluble fraction of TDP32 and mtRNP2, but not TDP35, using pan-TDP-43 antibody (Fig. 6C). These fragments were also detected using the anti-pTDP-43 antibody that reacts with phosphorylated serines at the C-terminus of TDP-43 (Ser409/410) (Fig. 6D). Therefore, our findings indicate that the 25 kDa CTF of TDP-43 is included in the aggregation of TDP-43 mutants lacking the RNP2, though TDP25 does not by itself form aggregates. Since GFP-tagged TDP25 has been reported to form aggregates [26][27][28], we assessed the solubility of GFP-TDP-43 fragments. GFP-TDP35, TDP32, and TDP25 were all intensely insoluble (Fig. S9A). However, the features of non-tagged TDP-43 fragments were similar to those of V5-TDP-43 fragments: TDP32 was substantially insoluble, whereas TDP35 and TDP25 were less insoluble (Fig. S9B). These findings suggest that V5 tag appears to be suitable to assess the solubility of TDP-43 fragments.

The 25 kDa CTFs Bind to TDP-43 Lacking the RNP2 Motif
Since 25 kDa and shorter CTFs were detected in the insoluble fraction of mtRNP2, we assumed that mtRNP2 binds to, and thereby sequesters, the 25 kDa CTF. To test this hypothesis, experiments using the 25 kDa and 14 kDa CTFs of TDP-43 were performed (Fig. 7A). We co-transfected V5-tagged TDP25 and GFP-tagged TDP-43 in HEK293 cells and fractionated the whole cell lysates to obtain RIPA-soluble and -insoluble fractions. The amount of V5-TDP25 in the insoluble fraction was remarkably increased by GFP-mtRNP2, although we hardly detected the band of V5-TDP25 in the insoluble fraction of the cells expressing wildtype TDP-43 (Fig. 7B). Immunocytochemical confocal microscope analysis also demonstrated that V5-TDP25 was colocalized with the aggregates of GFP-mtRNP2 in the cytosol (Fig. 7C).
Next we performed immunoprecipitation to examine the binding of small CTFs to mtRNP2. The results showed that TDP25 binds to V5-mtRNP2, but not to the V5-mtRNP2 (1-273) that lacks the C-terminal domain of TDP-43 (Fig. 7D, E). Since TDP25 includes the RRM2 (aa 191-262), a shorter CTF, TDP14, was also used to determine whether the RRM2 is necessary for the binding to mtRNP2. TDP14, which spans amino acids 274-414, lacks RRM2, but contains the C-terminal region where most ALSrelated mutations are located. The results showed that GFP-TDP14 efficiently binds to V5-mtRNP2 as well as GFP-TDP25 (Fig. 7F, G). In addition, we comfirmed that mtRNP2, as well as TDP35 and TDP32, showed no interaction with IgG/beads (Fig.  S10). These results suggest that small CTFs, seen in the neurons of ALS and FTLD patients bind to the C-terminal domain of mtRNP2 and are sequestrated into the cytosolic aggregates of mtRNP2, and that the RRM2 is not required for this interaction. Although our results suggest that wild-type TDP-43 also interacts with the CTFs, this might result from the effect of the GFP tag given that GFP-TDP25 tends to be insoluble as shown in Fig. S7A.

A Decrease in Cellular RNA Enhances the Insolubility of TDP-43
The RNP2 motif of RRM1 is important for its RNA binding ability [24]. To confirm that mtRNP2 lacks the ability to bind to RNA, we performed RNP immunoprecipitation. The results showed that the amount of RNA that binds to TDP-43 is decreased by the disruption of RNP2, although the efficiency of precipitation was similar between wild-type TDP-43 and mtRNP2 ( Fig. 8A-C). We also tested whether mtRNP2 binds to the 39UTR of the mRNA of human neurofilament light chain (NFL), a known target of TDP-43 [29,30], using RNP immunoprecipitation followed by PCR. The results showed that wild-type TDP-43, but not mtRNP2, binds to the 39UTR of NFL mRNA, confirming that mtRNP2 loses its ability to bind to the target RNA of TDP-43 (Fig. 8D).
Based on the observation that the disruption of the RNP2 motif increases the aggregation of TDP-43, we hypothesized that the decreased binding to RNA leads to the formation of insoluble aggregates of TDP-43. Therefore, we investigated the effects of RNase on the properties of the endogenous TDP-43. To increase the detection sensitivity of insolubilized TDP-43, we used a mild buffer and four-step fractionation. The results demonstrated that the amount of urea-insoluble endogenous TDP-43 was increased from that of the start sample (0 h) by a 16-h incubation with RNase (Fig. 8E).

Discussion
The ubiquitin-positive, phosphorylated inclusion of TDP-43in neuronal cytoplasm is a pathological hallmark of TDP-43 proteinopathy. Since this inclusion contains 18-26 kDa CTFs of TDP-43 that do not have the RRM1 [8,25], the phenotype of the TDP-43 CTFs have been intensively investigated. Previous studies showed that the 35 kDa TDP-43 CTF is sequestered into stress granules in cultured cells [31,32], whereas the 25 kDa CTF phenotype findings have been controversial [25][26][27][28]31,[33][34][35]: several reports did demonstrate that 25 kDa of TDP-43 forms cytosolic aggregates but most were tagged with fluorescent proteins. Therefore, a small tag, V5, was used to assess the cellular distribution of TDP-43 in our experiments.
We first focused on the physical features of two types of TDP-43 CTFs, TDP35 and TDP32. The cells expressing TDP32 only exhibited ubiquitin-positive and phosphorylated aggregations. In addition, the immunoblots also showed that TDP32, but not TDP35, was insoluble and phosphorylated, suggesting that the RNP2 motif in the RRM1 is responsible for the process of TDP-43 aggregation. Although the cells expressing TDP35 occasionally appeared to form cytoplasmic aggregates, those were the components of stress granules and were neither insoluble nor ubiquitinated. On the other hand, the cells expressing TDP25 did not form aggregates in our experimental conditions, although the TDP25 did not contain the RNP2 motif. This discrepancy might be explained by our observation that the disrupted RNP2 motif and the remaining RRM1, but not RRM2, are both required for the aggregations of TDP-43. In addition, the finding that DRRM1 TDP-43 does not form aggregates also supports this hypothesis.
Since the RNP2 motif in the RRM1 is responsible for the ability of TDP-43 to bind RNAs with specific sequences [24], a disruption of RNA binding could cause the aberrant aggregation of TDP-43. The cells with either DRNP2 or mtRNP2 TDP-43 formed aggregates and underwent both phosphorylation and ubiquitination. Disruption of RNP1, another RNA biding motif, also insolubilizes TDP-43. In addition, incubation with RNase caused insolubilization of endogenous TDP-43. These data further confirmed that RNA binding is important for the process of TDP-43 aggregation. Since the interaction of negatively charged RNA is responsible for the conformation of RNA-binding proteins, RNA may exert a chaperoning effect on its bound proteins [36]. Therefore, when the interaction of TDP-43 with RNA is disrupted, a consequent conformational change could cause TDP-43 to aggregate, and affect the function of NLS in this protein. Previous studies that demonstrated reduced levels of RNA in the motor neurons of ALS patients may support our hypothesis [37].
We have discussed the fact that the CTFs did not, except for TDP32, aggregate in normal conditions, whereas full-length TDP-43 could aggregate when RNA binding is disrupted. However, a question is raised as the major components of aggregated TDP-43 in TDP-43 proteinopathy patients are CTFs, such as TDP25. Although TDP25 did not by itself aggregate in our system, the cells expressing mtRNP2 or TDP32 contained ,25 kDa phosphorylated TDP-43 in their insoluble fractions. In addition, in the cells co-expressing mtRNP2 and TDP25, TDP25 formed aggregates and colocalized with mtRNP2, suggesting that mtRNP2 sequesters TDP25 in the aggregations.
On the other hand, mtRNP2 (1-273), which lacks the Cterminal domain, did not have the ability to sequester TDP25. The C-terminal domain of TDP-43, in which most of the disease mutations are located, contains a glutamine/asparagine-rich (Q/ N-rich) domain; also referred to as the prion-like domain, it is involved in the self-assembly of misfolded CTFs and the sequestration of TDP-43 into polyglutamine aggregates [38][39][40]. Therefore, it is possible that TDP-43 in which RNA binding is disrupted forms the initial aggregation core, and further sequestrates TDP-43 CTF into the aggregation through interactions with the C-terminal domain.
In summary, we demonstrated that the RNP2 motif in RRM1 plays a substantial role in pathological TDP-43 modifications and that disruption of RNA binding may underlie the process of TDP-43 aggregation. Figure S1 Immunocytochemistry of NSC34 cells expressing TDP35 or TDP32. Cells were stained with anti-V5 (red) and anti-TIAR (green) antibodies. The aggregates with TDP35, but not TDP32, colocalized with TIAR. Scale bar = 5 mm.
(TIF) Figure S5 The colocalization coefficient of the ubiquitin and V5 signals. Cells were stained with anti-V5 (green) and antiubiquitin (red) antibodies. Colocalization with ubiquitin was significantly higher in the cells expressing mtRNP2 than in those bearing wild-type TDP-43 (p,0.001). Scale bar = 5 mm. Error bars indicate SEM (n = 3). (TIF) Figure S6 The colocalization coefficient of the pTDP-43 and V5 signals. Cells were stained with anti-V5 (green) and anti-pTDP-43 (red) antibodies. Colocalization with pTDP-43 was significantly higher in the cells expressing mtRNP2 than in those with wild-type of TDP-43 (p,0.05). Scale bar = 5 mm.